PageWell
Sign in Get started

Docs

Everything an agent needs, in one fetch

Four entry points. The first two are for the machine doing the work; the last two are for whoever has to debug it at two in the morning.

SKILL

Install once, then the agent knows

The SKILL bundles the CLI, the directory conventions, and the two things the agent must not attempt. It is the shortest path from “my agent wrote some docs” to a URL.

$ pagewell skill install
copy into your agent's skills directory
CLI

Nine commands, no daemon

login · push · pull · share · ls · rm · token · usage · whoami. Every one of them prints JSON with --json, so the agent never parses a table.

$ pagewell push ./docs --json
the same output the SKILL reads
API

One contract, two repositories

api/openapi.yaml is the only contract. The CLI is generated against it; if the two ever disagree, the YAML is right. Errors carry a stable error.code enum that is additive-only and never translated.

$ GET /api/v1/spaces
bearer token · scoped
llms.txt

The site, flattened

Every share publishes one. It lists every page with its title, path and summary, so an agent on the other end reads the whole thing in a single request instead of crawling.

$ curl p.pagewell.ai/s/8xK2mQvR/llms.txt
no auth for public shares

Stable surface

What we promise not to break

An agent cannot read a changelog. These four are versioned promises, not implementation details:

Reading order

If you are wiring this up by hand rather than through the SKILL:

  1. 01Install the CLI and run pagewell login — device flow, approved in a browser.
  2. 02Run pagewell push ./docs --dry-run and read the plan before anything is written.
  3. 03Create a share: pagewell share ./docs --visibility code --expires 7d.
  4. 04Mint a scoped token for the agent: pagewell token create --scope docs:write.
  5. 05Point the agent at llms.txt, not at the HTML.